AI's next dataset is your apartment
This AI newsletter covers MicroAGI's Shift app offering free NYC apartment cleanings in exchange for POV video data to train robots, alongside news about Inherent Labs' self-improving science AI platform and various AI tool updates. The issue also features staff AI use cases, a tutorial on building a video workstation with Higgsfield and Claude, and a reader workflow using HubSpot's AI Prospecting Agent.
Summary
The newsletter's lead story covers MicroAGI's Shift app, which offers free home cleanings in New York City in exchange for AI training data. Cleaners wear head-mounted 'magic hat' cameras to record first-person footage of their work, which MicroAGI sells to AI labs and uses for its own robotics research. The service drew 'thousands and thousands of bookings' at launch, with expansion planned for London, Munich, and Zurich. The newsletter frames this as part of a broader trend — following DoorDash's similar initiative with couriers — where ordinary human labor is being converted into AI training data, with the subjects being simultaneously the customers, the workforce, and the training material for their own eventual automation.
The Roundtable section features staff sharing personal AI use cases: writer Zach used ChatGPT and Codex to plan and visualize a garden overhaul, generating a materials list, build instructions, and a companion planting map. Partnerships lead Joey uploaded body scan results and health data to Claude to generate a personalized workout and diet plan.
A training guide walks readers through building a short-form video workstation using Higgsfield and Claude Code, covering campaign creation, prompt generation, video output, and turning workflows into reusable skills. The guide recommends running manual campaigns before automating to refine skills from real feedback.
Inherent Labs, a London-based startup founded by ex-Google DeepMind researchers, emerged from stealth with $50M in funding. Their Faraday platform pairs scientists with self-improving AI agents designed to identify high-value research questions rather than just answering prompts. The startup applies recursive self-improvement logic not just to model training but to the entire research organization, including resource allocation and decision-making.
Additional news items include Microsoft merging its AI tools into a super app, SoftBank's $87B commitment to an AI data center in France, Nvidia's RTX Spark superchip for personal AI agents, OpenAI hiring for robotics, and OpenAI's Rosalind Biodefense platform for pandemic preparedness. A reader workflow highlights HubSpot's AI Prospecting Agent, which monitors buying signals and automates outreach sequences, enabling companies to reduce SDR headcount while improving lead quality.
Key Insights
- MicroAGI's Shift app demonstrates that human-captured POV footage from domestic labor is valuable enough to AI labs that a startup can offer free apartment cleanings and still turn a profit by selling the data.
- The newsletter argues that the next major AI dataset frontier is not the internet but ordinary human work — with Shift pushing this model uniquely deep into the home, where the same person is customer, worker, and unwitting robotics trainer.
- Inherent Labs' founders argue that self-improving AI should be applied to the entire scientific organization — including resource allocation and research prioritization — not just to model training loops.
- Reader Wyatt B. reports that HubSpot's AI Prospecting Agent has allowed his company to hire fewer SDRs by automating the identification of buying signals and drafting outreach emails, shifting human effort toward closing rather than cold prospecting.
- MicroAGI's Shift claims to have already paid out over $5M in Q1 to people worldwide filming everyday chores at $20/hour, suggesting the physical AI data collection economy is already operating at meaningful scale before the NYC consumer launch.
Topics
Full transcript available for MurmurCast members
Sign Up to Access